Machine learning fairness assessments improved by measuring uncertainty
The Role of Uncertainty in Assessing the Fairness of Machine Learning Models
Machine Learning
Summary
Machine learning models are used in important areas like healthcare and law enforcement, but sometimes they can be unfair to certain groups. To check fairness, it's not enough to look at just one model and its predictions. The authors talk about ways to measure how certain or uncertain we are about the fairness of different models. They explain methods from statistics that help us understand this uncertainty better using both made-up and real data. This approach helps make fairer decisions when choosing machine learning models.
machine learningfairnessbiasuncertainty quantificationrisk assessmentBayesian statisticsfrequentist statisticsmodel selectionpredictive accuracydisadvantaged groups
Authors
Francesca Panero, Ernst C. Wit, Marco Scutari
Abstract
Machine learning models are widely used in clinical applications, social media, law enforcement and critical infrastructure. Verifying whether their outputs are biased against disadvantaged groups or individuals is crucial to ensuring they are fair and allowing their use in such settings. A rigorous risk assessment of possible fairness violations requires quantifying the uncertainty associated with selecting and estimating such models. Yet, this is rarely done in the literature, which focuses on identifying a single model with a suitable trade-off between predictive accuracy and fairness. In this paper, we move beyond point estimation and discuss frequentist and Bayesian approaches to uncertainty quantification for fair machine learning, with practical examples and implications for simulated and real data.